Embed Size (px)
Citation preview
The Effect of Microfinance and its Impact on Economic Growth
- A Case Study of Rural Vietnam -
Pham Tien THANH1, Katsuhiro SAITO2 and Pham Bao DUONG3
No. 17-E-001
This Paper was presented at the International Conference on Economic Structure which was held on 17-19
March 2017, Meiji University, Tokyo, Japan. 1 Corresponding Author, School of Economics, University of Economics, Ho Chi Minh City 2 Department of Agricultural and Resource Economics, The University of Tokyo 3 Faculty of Economics and Rural Development, Vietnam National University of Agriculture
Abstract
Since its birth in 1970s, microfinance has been growing rapidly with the aim to reduce
poverty and to promote economic growth. In this paper, we focus on the effect of
microfinance in rural Viet Nam using the panel data of the Vietnam Access to
Resources Household Surveys (VARHS). We have two objectives: the first is to
examine the effect of microfinance at the micro level, and the second is to evaluate the
impact on economic growth. For the first objective, we employ the method for project
evaluation and identify the effect of microfinance on production activity and net
income of rural households who use microfinance. After examining the impact of
microfinance on poverty reduction, we aggregate the effect of production increase and
income growth among rural households and evaluate the macro impact with applying
input-output analysis. The results at micro level find that microcredit benefits self-
employment rather than other activities. The result from the macroeconomic
viewpoints is that effect of output increase is not so large.
Keywords: microfinance, poverty reduction, economic growth, input-output analysis,
VARHS
1
1. INTRODUCTION
Poverty reduction, access to education, clean water, sanitary, health care, etc.
are the top concerned among 17 Sustainable Development Goals - SDGs) which
formerly was Millennium Development Goals (MGDs). They are the important issues
and attract the attention of every country, especially the developing countries like
Vietnam. During the past decades, Vietnam has had remarkable achievements in the
socio-economic development. Vietnam is considered one of the few countries that
have obtained the remarkable achievement in poverty reduction. Statistics by World
Bank Indicators showed that the poverty rate (using GSO-WB Poverty Line) has
declined from 37.4 percent in 1998 to 17.2 percent in 2012 (Demombynes and Vu,
2015). The report also shows that the poverty rate in 2012 in rural areas (22.1 percent)
is four times higher than that in urban areas (5.4 percent). The statistics indicate that a
large number of rural households in Vietnam, especially the rural poor, still live in
poverty with under poor living standards and suffer from the lack of socio-economic
opportunities.
Poverty reduction, education, gender equality, good health and clean water and
sanitation, especially in the rural area, are the most concerned issues among the 17
Sustainable Development Goals. In the world and in Vietnam, many programs have
been implemented to achieve these goals.
Morduch and Haley (2002) state that credits can help the poor to improve their
living standards or at least cover their living expenses. However, a research by Brau
and Woller (2004) find that the poor have difficulties in accessing to formal credit
sources; particularly, the poor in the developing countries have more difficulties in
accessing than those in the developed countries. In Vietnam, many rural households
have difficulties in accessing to credit, especially poor households, households in
remote areas, the minority ethnic groups, or households operating in such fields with
high risk as aquaculture, etc., These households always have high demand for credit
(Duong and Izumida, 2002) but they have some difficulties in borrowing from semi-
formal and formal credit sources such as banks or financial institutions due to lack of
collateral assets. Thenceforth, many households have to borrow from informal credit
sources such as friends, relatives, money-lender, etc. The Government has taken a lot
2
of effort to support the rural households with accessing to semi-formal and formal
credit but the result is still rather limited. Microcredit is designed as a collateral-free
credit services and established to serve the poor, and thereby it can increase the poor
households’ accessibility to formal credit. Microcredit is defined as a small loan
granted to the poor so that they can run production or do business to generate income
and improve their well-being (Microcredit Summit, 1997).
Microcredit programs have applied in many countries as a tool of poverty
reduction and hunger eradication. Microcredit is considered as a tool for the goal of
fighting poverty and improving welfare via increasing their income or consumption
(Khandker, 1998; Yunus, 2003; Li et al., 2011; Khandker and Koolwal., 2016). Some
researchers argue that microcredit has no significant affect on households’ living
standards or only benefit the less poor households (Hulme and Mosley, 1996;
Coleman, 1999; Dupas and Robinson, 2008; Coleman, 2006). Takahashi et al. (2010)
conclude that microcredit has no significant impact on various outcomes, except for
sales from self-employment for the non-poor and schooling expense for the poor,
thereby indicating that microcredit has no immediate impact on poverty reduction.
In Vietnam, microcredit is found to improve households income, consumption
or self-employment profits, thereby reducing poverty-gap and contributing to the
poverty alleviation (Nguyen, 2008; Lensink and Pham, 2011; Phan et al., 2014). Reis
and Mollinga (2012) also conclude that microcredit program was founded to improve
the quality of water supply and the sanity system of the rural households in Vietnam.
However, neither of these studies investigates the role of microcredit in improving
different income sources.
Households are always rational when making decisions. Moreover, due to the
constraint in intra-household resources, the farmers may choose the appropriate
income-generating activities to optimize their benefit. Credit is fungible and thereby
credit borrowed for nonfarm activities can be diverted to other income-generating
activities such as agriculture (Khandker and Koolwal, 2016). Therefore, it is interested
and necessary to investigate the effect of microcredit on various income sources.
Literature document that few studies in Vietnam investigate the role of microcredit on
different types of income sources. Therefore, this research is conducted to examine the
3
role of microcredit in improving income from such activities as agriculture, self-
employment, common property resources, etc. To capture to better results, this
research applies match difference in difference (Match DID or PSM-DID) to
investigate the micro-impact of microcredit.
Though there are many studies which evaluate the impact of microfinance on
poverty and welfare at household level as explained above, surprisingly little is known
about macro impact. Within these studies, there are two types of studies, econometric
study and quantitative study such as computable general equilibrium analysis. For
example, Maksudova (2010) shows the Granger causality from microfinance to
economic growth with Microfinance Information Exchange (MIX) data. In contrast to
econometric study, Buera et al. (2012), Mahjabeen (2008) and Raihan et al. (2017)
employ computable general equilibrium model for evaluating aggregate and
distributional impact of microfinance targeted to small business, impact on household
consumption and welfare, and show the existence of macro impact. Essentially these
studies are counterfactual analysis, and are not based on the econometric study. Thus
we use the econometric results on the effect of microfinance at the household level,
and evaluate the macroeconomic effect, especially the increase in outputs, by using
input-output framework. To our best knowledge, there is no study for using input
output model for evaluating macroeconomic effect.
2. METHODOLOGY
2.1. Model for Analyzing Micro-Impact of Microcredit
2.1.1. Estimation Strategy
The objective of Impact evaluation of a program is to investigate the difference
in outcome between participation and non-participation in the programs. However, in
reality, we cannot observe one household at two stages at the same time. That is, there
is no household that can both participate and non-participate in a program. Impact
evaluation methods will construct a counterfactual to make comparison between
participating group (Treatment) and non-participating group (Control) with the most
similar characteristics. Thenceforth, it is possible to evaluate the impact of the
4
programs on the outcome (welfare). This research employs PSM-DID for estimating
the impact of microcredit on various welfare indicators.
Propensity Score Matching (PSM)
PSM method was initiated by Rosenbaum and Rubin (1983), and then it is
developed in many studies by Becker and Ichino (2002), Dehejia and Wahba (2002),
Khandker et al., (2010). On the basis of impact evaluation using PSM, the following
steps should be conducted:
Firstly, a probit model is conducted to estimate determinants of accessibility to
microcredit. The estimated probability of participation (or propensity score) of each
household in the research data is calculated from this model. The equation is written
as follows:
Pr(Cri=1) = β0 + β1Zi
Where, Cri denote the participation in microcredit program (1=Borrow;
0=Non-borrow). Zi represents determinants of the accessibility to microcredit.
In the next step, the common support region is specified. In this step, some
observations of control group may be dropped out because they have too high or two
low estimated probability. Also in this step is the balancing test conducted to testify
whether, in each block, the average Propensity score and mean of X are not different
between treated units and control units. Dehejia and Wahba (2002) state that this test
is conducted via distributing the observations into blocks based on the estimated
propensity scores.
At the matching step, each treated unit is matched with one or some control
units based on the most similar propensity score. Then the outcomes between each
treated unit and control units are compared. The difference from this comparison
reflects the impact of microcredit programs with respect to each treated unit, or
individual gain. In order to match these two groups, various techniques of matching
may be applied such as Nearest-neighbor, Caliper (or radius), Stratification (or
interval), and Kernel matching.
Average outcome of all individual differences is then calculated to capture
overall mean value that is considered as impact of microcredit program.
5
PSM has been a common method for policy researches including impact
evaluation of credit programs. PSM may to reduce the selection biases when
estimating the results. However, this method results in some limitations such as: (1)
PSM does not measure the difference in an outcome pre–post attendance in the
program overtime. (2) Hidden bias may still exist because PSM do not include
unobservable characteristics.
Difference in Difference (DID)
Followed Lester (1946) and Khandker et al. (2010), the model using DID is as
follows:
itit CRTCRTY *3210
Where, Yit denotes outcomes of households i at time t. Cr is treatment status
(1=Treatment; 0=Control) or Participation in microcredit programs (1=Borrow;
0=Non-borrow). T denotes time Variable (0=Before Treatment; 1=After Treatment).
𝜺𝒊𝒕 is error term.
(�̂�0) and (�̂�0 + �̂�1) are the mean outcomes of control group before and after
program, respectively. Meanwhile, (�̂�0 + �̂�2) and (�̂�0 + �̂�1 + �̂�2 + �̂�3) are the mean
outcome of treatment group before and after program, respectively. Accordingly, (�̂�2)
and (�̂�2 + �̂�3) are the single differences between two comparison groups before and
after program, respectively. The DID estimate is the variation in outcome between
two comparison groups before and after program. Therefore, ( �̂�3 ) is estimated
coefficients using DID.
PSM-DID
PSM-DID is a combination of PSM and DID using panel data. Khandker et al.
(2010) state that PSM-DID can capture better results due to its reduction in estimation
bias. Based on PSM and DID methods, the PSM-DID include some main stages
including (1) calculation of propensity score; (2) balancing test and common support ;
(3) DID combined with matching to match treatment with control group and estimate
the impact of program. In this research, the PSM-DID is estimated using command
diff developed by Villa (2016). Command diff combines DID estimation with kernel
6
matching (Heckman et al., 1997, 1998; Blundell and Dias, 2009; cited in Villa, 2016).
The kernel weights are also incorporated to capture a kernel matching DID treatment
effect.
2.1.2. Selection of variables
Microcredit, in this research, is defined the small loan loans from formal and
semi-formal sources that are used for income-generating activities such as production
or self-employment. As prescribed in the Decision No. 306/QD-TTg of the Prime
Minister, the maximum amount that Vietnam Bank for Social Policy can grant the
poor is 100 million dong. Moreover, Khoi et al.’s (2013) research in Vietnam specifies
the amount of microcredit no larger than 100 million dong. On these basis, this
research also limits the loan amount under 100 million dong to be considered as
microcredit.
The independent variable used for calculating Propensity Score are presented at
Table 2. There is no firm theory on the selection of variables to be incorporated into
the model of determinants of accessibility to credit sources, including formal or
informal. The empirical evidences document that factors affecting households’
probability of accessing microcredit may includes characteristics at household head
level, household level, region level and institutional level (Duong and Izumida, 2002;
Li et al, 2011; Phan et al., 2013; Li et al., 2013). On the basis of literature review and
data availability, this research selects the independent variables (see Table 2) to
include in the research model. Based on the research by Khandker et al. (2016),
Takahashi et al. (2010), Lensink and Pham (2012), the outcome variables (dependent
variables) used for estimation the micro-impact of microcredit are presented in Table
3.
2.2. Model for Analyzing Macro-Impact of Microcredit
The fundamental system of equations are written as follows:
)(ˆ
)(ˆ
222212122
121211111
2222221212
1112121111
FXAXAMM
FXAXAMM
MEFXAXAX
MEFXAXAX
7
where iX are sub vector of output column vector, ijA are sub matrix of input-
output coefficient matrix, iF are sub-vector of final demand, iE are sub vector of
export, iM are sub vector of import andiM̂ are diagonal matrix of import to domestic
demand ratios. First two equations are demand and supply equilibrium conditions,
while third and fourth equations show that the import is determined as a portion of
domestic demand.
In the standard input-output model, final demand vectors are assumed to be
exogenous, and we obtain output by the following system of equations:
222
111
1
22222122
121111111
2
1
)ˆ(
)ˆ(
)ˆ()ˆ(
)ˆ()ˆ(
EFMI
EFMI
AMIIAMI
AMIAMII
X
X
The associated comparative static version of the above model is
222
111
1
22222122
121111111
2
1
)ˆ(
)ˆ(
)ˆ()ˆ(
)ˆ()ˆ(
dEdFMI
dEdFMI
AMIIAMI
AMIAMII
dX
dX .
When we evaluate the macroeconomic effect of micro finance, we treat the
output of some industries as exogenous. Suppose 2X as exogenous. First and third
fundamental equations are used for obtaining the following:
11112121
1
1111 )ˆ()ˆ()ˆ()ˆ( EMIFMIXAMIAMIIX
This is one of the modifications in this paper. We can evaluate the endogenous
output change due to the change in exogenous output by
2121
1
1111 )ˆ()ˆ( XdAMIAMIIdX
.
From the second fundamental equation, 2M is determined as well. This formula
examines only interindustrial linkage due to the intermediate transaction among
industries. Another modification is treating final demand as endogenous. Since when
output of some industries change, value added is changed as well. This induces the
change in income of laborers for example, and as a consequence final demand will be
changed. The final demand function is modified as follows:
)( 2211111011101 XVXVfcFFFF
)( 2211212021202 XVXVfcFFFF
8
where iV are row vector of value added coefficients, 1c is marginal propensity to
consume, and if is share vector of final demand. In this formulation, we obtain
122111121
1
11111111 ))ˆ()ˆ())ˆ()ˆ(( EXVfMIcAMIVfMIcAMIIX
.
Comparative static version is
22111121
1
11111111 ))ˆ()ˆ(())ˆ()ˆ(( dXVfMIcAMIVfMIcAMIIdX
.
Using this formula, we can evaluate the income linkage as well as
interindustrial linkage. That is, the effect of final demand change on industrial output
due to income change is captured.
3. RESEARCH DATA
3.1. Data Source
VARHS are conducted under the cooperation of the Central Institute for
Economic Management (CIEM), Ministry of Planning and Investment (MPI), the
Center for Agricultural Policy (CAP), Institute of Labor Science and Social Affairs
(ILSSA), Ministry of labor - Invalids and social affairs (MOLISA); the Development
Economics Research Group (DERG), the University of Copenhagen; and the Ministry
of Foreign Affairs (DANIDA), Denmark. Vietnam Access to Resource Household
Survey (VARHS) is a large-scale survey. This survey collects data from 3703 rural
households in 47 communes located in 12 provinces in Vietnam, including Ha Tay,
Lao Cai, Phu Tho, Lai Chau, Dien Bien, Nghe An, Quang Nam, Khanh Hoa, Dak Lak,
Dak Nong, Lam Dong and Long An. These 12 provinces represents 7 socio-economic
regions in Vietnam, including Red River Delta, North East, North West, North Central,
South Central Coast, Central Highlands and Mekong River Delta.
VARHS survey provides detailed information about a wide range of important
demographic, economic and social characteristics of the farm households, such as on
farm- and farmer-specific attributes, resources endowment, agricultural inputs and
outputs, economic activities and welfare, savings and borrowings etc., From 3703
households survey in 2012, 3644 households are re-interviewed in 2014. In order to
create a balance panel data, some observations with missing data are dropped out of
9
the research sample. The final sample used for estimation are 7088 observations,
including households each survey.
For estimating the macro impact of microcredit, we also employ Vietnamese
Input-Output (IO) table of 2007 published by the General Statistics
Office of Vietnam (GSO). This IO table provides a useful framework that accounts for
interrelationships among 138 sectors in Vietnam’s economy. Input-Output (IO) is
considered as a useful tool for analyzing the relationship between microcredit and
economic activities, thereby capturing the impact of microcredit on economic growth.
3.2. Descriptive Statistics
Table 1 reports the loans obtained by households in this research. Table 1
show that for the case of pooled sample, the number of households granted with loans
is 2654, accounting for 37.44%. The number of borrowers tend to reduce over time
while the average amount increases from 2012 to 2014, which indicates that credit
provider seem to give priority to the amount of each loan rather than the number of
loan. This trend is similar for the case of microcredit and loan form production
(formal or informal). However, for the case of microcredit and informal loan for
consumption, there is an inverse trend; that is, the number of borrowers tend to
increase but the loan amount reduces over time. For the case of formal and semi-
formal loan for consumption, both number of borrowers and loan amount tend to
increase over time. Table 1 also report that among 2654 borrowers, there are 1908
borrowers from formal and semi-formal sources and 967 from informal sources,
which indicates that some households borrow from both sources. The total amount of
informal loan imply that informal credit sector still plays a significant role in rural
financial market in Vietnam. This information is similar to Barslund and Tarp’s
(2008) research in Vietnam, in which find that credit informal sector co-exist with
formal sector and accounting for about one-third of all loans. The explanation is that
the rural poor households still rely on informal networks and relatives.
10
Table 1 – Loans Obtained
Number of Borrowers Average Amount
2012 2014 Pooled
2012 2014 Pooled
ANY LOAN 1408 1246 2654
40.474 57.248 48.349
Microcredit 572 313 885
27.845 35.856 30.678
Non-Microcredit 922 974 1,896
63.558 50.013 56.971
BOTH 86 41 127
Formal & Semi-formal 1,017 891 1,908
42.578 67.138 54.047
Production 637 362 999
50.055 77.133 59.867
Consumption 449 583 1,032
30.540 61.467 48.011
Informal 512 455 967
37.798 34.523 36.257
Production 350 192 542
36.192 38.671 37.070
Consumption 217 308 525 37.961 31.971 34.447
BOTH 121 100 221 Note: Average amount in Million VND
885 households are found to borrow from microcredit sources and of those,
there are 572 borrowers in 2012 and 313 borrowers in 2014. Meanwhile, there are
1896 borrowers from other non-microcredit sources, and of those 922 borrowers in
2012 and 974 borrowers in 2014. The results on number of borrowers from any source,
microcredit and non-microcredit sources indicate that some households borrow from
both sources. Similarly, the results on number of households borrowing from any
source, formal and semi-formal source and informal sources show that some
households have access to both sources.
The objective of this research is to evaluate the impact of microcredit program
using PSM-DID. Therefore, the households who borrow from microcredit at T=1 are
considered as treatment; the remaining households are control. Therefore, in this
research, there are 313 microcredit borrowers at T=1, which indicates that the number
of treatments are 313 households (616 observations in both surveys). The descriptive
statistics present the characteristics of treatment and control groups over times.
Table 2 shows some statistical summary on the characteristics of treatment and
control group in the first wave (baseline or year 2012 or T=0) and second wave
(follow-up or year 2014 or T=1). There is significant difference between treatment
and control groups in terms of some characteristics.
11
Table 3 shows the mean of outcomes of both group as well as the difference in
outcomes between them at both waves. For most of the cases, the borrowing group
seem to have higher outcomes than their non-borrowing counterparts, but only some
of the difference in these indicators are statistically significant. Most of average
outcomes of both groups tend to increase over time. For instance, treatment group
appear to have significantly higher total output value and income from agricultural
activities than control group at both waves. Further investigations on sub-sector of
agriculture show that treatment group have greater total output value from crop
production (significant in both waves) but there is no significant difference between
these groups in terms of income from crop production. Treatment group have
significantly higher total output and income from livestock than control group at
baseline, but there is no difference at follow-up. There is no significant difference
between two groups in terms of output value or income from self-employment and
common property resources. Treatment group appear to have lower wage income than
control groups and this difference is significant at both waves. There is no significant
difference between the two groups in terms of total output value and income from
earned sources (including and excluding wage income).
12
Table 2 - Variables for Estimation of Participation in Microcredit
Variable T=0 T=1
Treated Control Diff
Treated Control Diff
Education of HHH (Grade) 6.21 6.30 -0.093
6.68 6.56 0.128
Age of HHH 47.22 49.76 -2.54**
48.63 51.46 -2.835***
Marital Status of HHH
(1=Married)
0.86 0.83 0.032 0.87 0.82 0.056*
Gender of HHH (1=Male) 0.86 0.82 0.0408
0.85 0.80 0.053*
Ethnic of HHH (1=Kinh) 0.58 0.66 -0.081**
0.58 0.66 -0.075**
Microcredit (1=Yes) 0.29 0.15 0.145***
Non-Microcredit (1=Yes) 0.32 0.25 0.062*
0.13 0.29 -0.158***
Poverty Status (1=Yes) 0.27 0.25 0.012
0.20 0.19 0.016
Saving Value (Mil dong) 17.40 27.83 -10.43
20.22 30.75 -10.53*
Agricultural Land (hectare) 1.03 0.76 0.266***
1.03 0.73 0.309***
Residential Land (hectare) 0.14 0.10 0.043**
0.13 0.10 0.035*
Household Size 4.89 4.47 0.427***
4.86 4.41 0.451***
Dependence Ratio 0.30 0.35 -0.051**
0.29 0.36 -0.068***
Agricultural Labor 3.05 2.45 0.597***
3.19 2.49 0.704***
Self-employment Labor 1.09 1.02 0.0735
1.18 1.20 -0.024
Wage Labor 0.41 0.45 -0.042
0.44 0.37 0.068
Distance to main road (km) 2.03 2.15 -0.115
1.80 1.86 -0.054
Social Capital 7.27 7.43 -0.156
7.60 7.36 0.246
Poor Commune 0.61 0.51 0.095**
0.48 0.38 0.095**
Market (1=Yes) 0.52 0.57 -0.043
0.63 0.67 -0.049
Red River Delta (1=Yes) 0.13 0.17 -0.038
0.13 0.17 -0.038
North East (1=Yes) 0.16 0.19 -0.030
0.16 0.19 -0.030
North West (1=Yes) 0.21 0.17 0.033
0.21 0.17 0.033
North Central (1=Yes) 0.04 0.07 -0.032*
0.04 0.07 -0.032*
South Central Coast (1=Yes) 0.04 0.13 -0.096***
0.04 0.13 -0.096***
Central Highlands (1=Yes) 0.34 0.19 0.150***
0.34 0.19 0.150***
Mekong River Delta (1=Yes) 0.10 0.09 0.012
0.10 0.09 0.012
Obs 313 3231 313 3231
Note: Difference = Mean (Treatment) - Mean (Control)
Continuous variables are tested using ttest; Dummies in Italic at the bottom are tested using
prtest.
*, ** and *** : Significant at 10%, 5% and 1%, respectively.
13
Table 3 – Outcome Variables for Estimation of Micro-Impact of Microcredit
T=0
T=1
Variables Treated Control Diff
Treated Control Diff
INCOME
Agriculture 33.950 23.411 10.54*** 35.030 26.935 8.094*
Crop 4.172 4.311 -0.139 2.925 5.116 -2.191
Rice 10.486 6.581 3.905*** 11.054 7.099 3.955**
Maize 2.012 1.616 0.396 2.278 1.460 0.817*
Livestock + Aqua 13.060 6.776 6.284** 10.571 8.076 2.496
Livestock 8.323 4.735 3.588** 4.839 5.295 -0.456
Aquaculture 4.737 2.040 2.696 5.732 2.781 2.951
Self-employment 7.534 15.085 -7.551 16.051 16.488 -0.437
Common property
resources
2.002 1.614 0.388 1.925 1.885 0.041
Wage 16.530 22.971 -6.441*** 23.464 30.437 -6.973*
Total Earned Sources 60.016 63.081 -3.065 76.470 75.745 0.725
TOTAL OUTPUT VALUE
Agriculture 72.048 47.714 24.33*** 79.293 53.576 25.72***
Crop 45.528 32.011 13.52** 52.359 35.133 17.23***
Rice 18.678 11.524 7.154*** 19.868 12.435 7.433***
Maize 2.840 2.328 0.511 3.607 2.342 1.265*
Livestock + Aqua 26.519 15.703 10.82** 26.934 18.443 8.491
Livestock 22.682 14.224 8.457* 22.628 16.433 6.195
Aquaculture 3.838 1.479 2.358* 4.306 2.011 2.296
Self-employment 30.019 69.311 -39.29 60.543 70.953 -10.41
Common Property
Resources
2.315 1.715 0.600 2.078 2.125 -0.047
Wage 16.530 22.971 -6.441*** 23.464 30.437 -6.973*
Total Earned Sources 120.912 141.712 -20.80 165.378 157.092 8.286
Obs 313 3231 313 3231
Note: Difference = Mean (Treatment) - Mean (Control)
The dummies in Italic. ; Unit in Million VND
*, ** and *** : Significant at 10%, 5% and 1%, respectively.
4. RESULTS AND DISCUSSIONS
4.1. Participation in Microcredit Program (Propensity Score) and Balancing Test
Table 4 shows the estimation results on determinants of accessibility to
microcredit using probit model as the first stage of estimation using PSM-DID. These
results are used for the calculation of propensity score. Khandker et al. (2010)
suggests that the independent variables used for estimating the probability of
participation in a program should be in T=0. Therefore, most of characteristics for
14
estimating propensity score as T=0, except for Non-microcredit, which include the
information in T=0 and T=1. The estimations using probit, pscore or diff give the
same results. Therefore, we only report one estimated result. Max VIF equals to 4.32,
which indicates that there is no multi-colinearity in this model. The result on
balancing test is satisfied.
The results on common support regions specify that 296 households (using diff
command) or 302 households (using pscore) fall in off-support region. A further
investigation reports that difference using these commands is 8 households. Due to
the second stage using diff to estimate the average impact of microcredit, 6792
households in common support regions specified by diff will be used for analysis.
4.2. Micro Impact of Microcredit Program
Table 5 show that total income from earned sources (agriculture, self-
employment, common property resources and wage) seem to be unchanged when
households have access to microcredit. This finding is somewhat similar to Takahashi
et al. (2010) and Phan et al. (2014), who find no role of microcredit in improving total
income, but contrary to Khandker and Koolwal (2016) and Li et al. (2011), who
conclude that microcredit significantly increase total earned income. However, the
total production value from earned sources are found to increase when households can
borrow from microcredit sources and this effect is significant at 5 percent. A plausible
explanation is that households can simply raise output via increasing input, but they
can not gain the optimal input mix to improve raise profits (Takahashi et al., 2010)
15
Table 4 – Probit Estimations on Determinants of Accessibility to microcredit.
Variable Coef. t-stat
Microcredit (1=Yes) 0.4612*** 5.77
Non-Microcredit at T=1 (1=Yes) -0.6867*** -8.04
Non-Microcredit at T=0 (1=Yes) 0.1881** 2.54
Education level of HHH (Grade) 0.0066 0.63
Age of HHH -0.0079*** -2.69
Marital Status of HHH (1=Married) -0.1677 -1.32
Gender of HHH (1=Male) 0.0542 0.44
Ethnic of HHH (1=Kinh) 0.1068 1.01
Poverty Status (1=Yes) 0.0195 0.24
Saving Value (Mil dong) -0.0020** -2.37
Agricultural Land (hectare) -0.0101 -0.18
Residential Land (hectare) 0.1523 1.07
Total Land (hectare) 0.0211 0.44
Household Size 0.0094 0.37
Dependence Ratio -0.2620* -1.79
Agricultural Labor 0.1123*** 3.59
Non-Wage (Non-farm) Labor -0.0249 -0.74
Wage Labor -0.0360 -0.9
Distance to main road (kilometer) -0.0183* -1.89
Social Capital 0.0088 1.52
Poor Commune 0.0004 0.01
Market (1=Yes) 0.0675 0.91
Mekong River Delta (Base)
Red River Delta (1=Yes) -0.3074** -2.08
North East (1=Yes) -0.3874** -2.57
North West (1=Yes) -0.1281 -0.81
North Central (1=Yes) -0.5288*** -2.72
South Central Coast (1=Yes) -0.7818 -4.4
Central Highlands (1=Yes) 0.0288 0.21
Constant -1.0333*** -4.00
Max VIF 4.53
Balancing test Satisfied
Off-support 296 [302]
On-Support 6792 [6786]
Note: The dummies in Italic.
*, ** and *** : Significant at 10%, 5% and 1%, respectively.
Common support region is identified using such commands as diff and pscore [in bracket].
16
Moreover, rural households may choose to diversify their income sources and
may concentrate in one or some main activities to do investment to optimize the return.
Moreover, microcredit may be effective in some sectors at disaggregate but not
aggregate level. Therefore, the next section will investigate the impact of microcredit
on sub-categories of earned income sources, including agriculture, self-employment
and common property resources.
Table 5 – Micro-Impact of Microcredit on Income and Total Output Value by
Main Sectors
VARIABLE Total Production Value Income
Coef. t-stat
Coef. t-stat
Total Earned Sources 24.279** 2.00
1.991 0.54
Agriculture 0.157 0.03
-3.612 -1.35
Crop 1.935 0.53
-3.378** -2.34
Rice -0.288 -0.15
-0.354 -0.32
Maize 0.672 1.21
0.397 1.06
Livestock + Aqua -1.778 -0.47
-2.737 -1.09
Livestock -2.452 -0.72
-3.989*** -2.94
Aquaculture 0.674 0.43
1.252 0.60
Self-employment 25.444** 2.25
6.488*** 2.86
Common Property Resources -0.698** -2.04
-0.453 -1.53
Wage -0.433 -0.25 -0.433 -0.25
Note: *, ** and *** : Significant at 10%, 5% and 1%, respectively.
Regarding agricultural sector, the results show that microcredit seems to reduce
income and increase total output value ; however, neither of these effects are
statistically significant. This is quite consistent with Takahashi et al. (2009), who find
that microcredit does not improve sale or profit from agricultural activities. More
specifically, Takahashi et al.’s (2010) findings show that the effect of microcredit on
these outcomes is negative but statistically insignificant. However, the findings in this
research are different with Khandker and Faruqee’s (2003) in Pakistan, in which
conclude that the impact of credit on net value from agricultural activities are
significantly positive.
However, Karlan and Goldberg (2007) state that microcredit may have no
impact on outcomes in short term, for instance, one year. Because some crops or
17
animals take time to gain the returns, it is better to divide agricultural sectors into sub-
sector to investigate in details.
When investigating further on some sub-sectors of agricultural, microcredit is
found to have no impact on improvement of income and total output value from
raising animal, including livestock and aquaculture. However, further investigation
find that microcredit have negative impact on income from raising livestock. That
may be because some livestock consume high investment in inputs (including
Production loan interest payment) but take time to gain returns (for instance, cow,
buffalo, etc.). Thenceforth, the borrowing households can not benefit from raising
livestock in short term.
Meanwhile, microcredit is found to reduce income from crop, which is in
contrast with hypothesis. A plausible explanation for the decrease is that households
may shift their income-generating activities from crop production to other activities
such as raising livestock, aquaculture or self-employment when they borrow from
microcredit sources. This may be proved via no significant change in cost or
production value when households have access to microcredit.
The coefficients of net income and production for some main annual crops such
as rice or maize are found to be insignificant, which indicates no role of microcredit in
production of rice and maize. The first reason is that for the farmers who cultivate rice
or maize, these agricultural products seem to be a long-established and traditional
cultivation activity. Therefore, no matter whether they have access to microcredit,
they may still continue to cultivate these products.
Moreover, in order to improve output from these products, new agricultural
technology should be invested, which incur great cost and high risk. Therefore, it may
explain why microcredit, characterized by a small amount, plays no role in the
improvement of income from rice or maize. It is interesting to investigate further on
the impact of microcredit on each sub-sectors of crop production, but there is lack of
information on production cost of the other crops in detail. However, it may be
inferred that microcredit leads to a reduction in income from crops other than rice and
maize. Another plausible explanation is that crop production is riskier than livestock
production since climate shocks (i.e. flood, drought) seem to affect crop rather than
18
livestock production (Vilhelm et al., 2015). Thenceforth, the output from crop
production may not yield the high outputs.
In summary, for agricultural activities, including crop and livestock production,
another conceivable explanation is that the investment using microcredit may be not
effectives due to difference in agricultural skills across households, thereby affecting
the output to crop and livestock agriculture (Dearcon, 1998). The findings are
inconsistent with Khandker and Koolwal’s (2016), who found that microcredit has no
impact on crop income but significantly positive effect on livestock income.
Microcredit is found to have no impact on the difference in income from
common property resources; meanwhile it is found that microcredit borrowers have
lower total output value than the non-borrowers and this effect is significant at 5
percent level. The lower total output can be simply explained by the less investment in
inputs. It is somewhat in contrast with the statement that credit finances deforestation
(Ozorio de Almeida and Campari, 1995; Barbier and Burgess, 1996; Andersen, 1997;
Pfaff 1997; cited in Angelsen and Kaimowitz, 1999). Meanwhile, this finding is
similar to the works by Godoy et al. (1997) who find that families with credit may be
less dependent on forest-based activities or may choose to invest in off-farm activities.
In other words, access to credit may reduce exploitation of common property
resources such as forest clearance. The poor households seem to be greatly dependent
on common property resources such as pasture or forests (Jodha, 1992). In this
research, a large proportion of the rural poor (around 56.47%) depend on common
property resources for generate income. With access to credit, the rural households,
especially the poor, may have more opportunities to get a more decent jobs via non-
farm activities as suggested by Godoy et al. (1997), thereby being less dependent on
common property resources. Another explanation is that borrowers may commit to
comply with environmental requirement as a condition of rural credit (Assunçãoa et
al., 2013).
As strongly expected, microcredit has strong positive effect on self-
employment income and total output value. The result is quite consistent with
Khandker and Koolwal’s (2016) research in Bangladesh; specifically microcredit is
found to improve self-employment income because this is the original purpose of
19
microcredit program. A research in Vietnam by Lensink and Pham (2012) also find
that microcredit truly improves self-employment profit.
However, the estimated results in this research are partly different with the
findings in Indonesia by Takahashi et al. (2010), who find that microcredit only have
improved sale of self-employment but have no impact of self-employment profit. In
Takahashi et al.’s (2010) research, the estimation is also conducted using PSM;
however, the time length between two surveys is short (one year). Thenceforth, that
may partly explain why microcredit only contributes to the expansion of self-
employment but not improve the profit from this activity within one year.
As expected, microcredit is found to have no significant effect on households’
wage income. There is no firm theory on the relationship between microcredit and
wage income activities in rural areas.
For the estimation of macro impact of microcredit, this section also briefly
reports some results on the effect of microcredit on Total Output Value of sub-sectors
categorized based on Vietnam Input-Output table 2007.
For the case of agriculture, aquaculture and forestry, microcredit is found to
have significant impact on some sectors. In particular, microcredit is significant and
positively associated with the increase in total output value of sugarcane, other
livestock, and aquaculture. Meanwhile, microcredit is found to have negative impact
on total output value of timber. Microcredit seems to increase total output value of
such sector as other annual crop, rubber, coffee, tea, cow and buffalo and other
forestry activities; however, these effects are not statistically significant. Similarly,
there is no evidence to conclude that microcredit has negative impact on such sectors
as other perennial crop, pig, or poultry.
20
Table 6 – Micro-Impact of Microcredit on Total Output Value by Sub-sectors of
Agriculture, Forestry and Aquaculture
VARIABLE Coef. t-stat
Paddy -0.288 -0.15 Sugarcane 1.178** 2.00
Other Annual Crop 0.752 0.92
Rubber 0.206 0.47
Coffee 0.278 0.11
Tea 0.192 1.16
Other Perennial Crop -0.472 -0.33
Cow, Buffalo 0.108 0.12
Pig -2.882 -1.21
Poultry -0.303 -0.40
Other Livestock 2.203** 2.28
Aquaculture 4.999** 2.13
Timber -1.203*** -4.44
Other Forestry 3.601 1.13
Note: *, ** and *** : Significant at 10%, 5% and 1%, respectively.
Table 7 – Micro-Impact of Microcredit on Total Output Value by Sub-sectors of
Self-Employment
VARIABLE Coef. t-stat
Case 1
Manufacture of Beverage 0.344 1.138
Manufacture of Fabricated Metal Product 0.059 0.092
Manufacture of Furniture 5.406 1.607
Repair 0.293 0.891
Wholesale, Repair of Motor Vehicles 1.391 0.546
Wholesales 9.104*** 3.118
Retail 0.15 0.024
Food and Beverage Service -0.312 -0.190
Other Services 3.092* 1.672
Case 2
Manufacture of Food products & Beverage 2.548 1.216
Manufacture of Wood & Furniture 6.439* 1.770
Wholesale & Retail, Repair of Motor Vehicles 10.645 1.381
Case 3
Wood & Furniture, Paper, Printing 6.369* 1.743
Other Services -1.346 -0.390
Note: *, ** and *** : Significant at 10%, 5% and 1%, respectively.
21
For self-employment activities, microcredit is found to have positive impact on
most of sub-sectors but its effect is significant for some sector including wholesales,
Manufacture of wood and furniture (both including and excluding paper and printing)
and other services (Case 1). Microcredit seems to reduce total output value of some
sectors such as Food and beverage service and other services (Case 3), but these
effects are not significant.
4.3. Macro-Impact of Microcredit Program
As stated in the introduction, we have two aims: the examination of the effect
of microfinance at the micro level and the evaluation of the effect on economic growth.
This section is devoted to the second objective. In the previous sections, we have
examined the effect of microfinance at the household level with using the method of
project evaluation and found that the output values of some production activities (for
instance, wholesale) are increased significantly by microcredit.
Table 8 shows the output change in significantly affected sectors. Since the
coefficient is the average change in output value, we multiply the estimated number of
household which are using microfinance.
What is the macroeconomic consequence of these output change? We use
modified input output model, since it is easier to evaluate the result from econometric
study conducted in previous sections. And this is, to our best of knowledge, the first
attempt for evaluating macroeconomic effect of microfinance with input output
framework. Vietnamese Input output table of 2007 is used. This table contains 138
sectors
Exogenous sectors4 whose outputs are set exogenous are shown in Table 8.
Other sectors are set as endogenous sectors. The marginal propensity to consume is
assumed to 0.7. The model used for evaluation macroeconomic impact of
microfinance is explained in previous section. Table 9 shows the macroeconomic
impact of microfinance on output value and Value added. Exogenous output increase
is 180,262,332 million VND, and induced output increase is 4,670,194 million VND.
4 Other service in Table 8 is not considered as an exogenous sector, since it is very hard to identify the
corresponding activities in sectors in Input Output Table.
22
Total increase in output value is 8,896,714 million VND, which is 0.31% of total
output value. It is small in percentage. Since the direct effect of microfinance
(exogenous output change) is 4,226,512 million VND, the indirect effect is about
115% of direct effect; that is, 115% of direct effect is spillover effect generated by the
interindustrial linkage. If we take into consider the income linkage, total output
increase is 10,655,906 million VND. It is 1.20 times of the output increase without
income linkage. The effect through income flow is 1,759,192 million VND (41.6% of
direct effect).
Table 10 shows the macroeconomic effect of microfinance by four main
industrial classifications. Relatively larger effect is on agriculture, forestry and fishery
sector. Though mining sector is considered as an independent sector of microfinance,
its output increases. The per cent change in output of this sector is only 0.05% and is
relatively small comparing to other classifications. This is because of the
interindustrial linkage. More interesting is that the ratio of output increase with
income linkage to that without income linkage is 1.38 which is second following to
service sector. The interindustrial linkage and income linkage cannot be captured by
the micro-econometric method. Even if the effect of microfinance on income is
negative which is reported for example in Table 5, it might be positive if we take into
account the interindustrial linkage and income linkage5.
Agricultural sector includes many poor farmers, and we would infer
microfinance might boost the income of poor farmers depending on the magnitude of
linkage effect (a kind of trickle down effect).
Though large macroeconomic impact of microfinance is reported in literatures
using CGE model, our input-output model shows relatively small macroeconomic
impact of microfinance. One reason that the macroeconomic impact is small is that
the data set we used in econometric study includes only rural household. If we take
into account urban households, the macroeconomic impact is surely larger. In addition,
the productivity increase in production sector and financial sector is not taken into
account in our model. Only the effect of output increase is considered. Our model
5 Value added in a sector is assumed to be proportional to the output level in Input-Output Analysis. Thus the
value added of the sector in which output value increases is “by definition” increases. This fact may be
contradict to the findings we got in previous sections.
23
evaluates the short run effect. Other reason for small macroeconomic impact is that we
select only industries with statistically significant effects as exogenous sector.
Since the output change in exogenous sector is estimated as the products of
coefficients from project evaluation estimated in previous section and the number of
household using microfinance. Thus the exogenous output change depends on the
value of coefficients and the estimated number of household using microfinance. The
value of coefficients which is an average output change in each production activity
depends on the measurement period of the estimation (it takes longer period to get full
effect of investment via microfinance), level of the production and entrepreneur skill
of the farmers who use microfinance, and the amount of microfinance. The number of
household using microfinance depends on the transaction cost and procedure for
acquiring microfinance, attitude toward risks of borrowers and opportunity for other
source of credits.
The target client of microfinance programs are typically the poor or the near-
poor households. These households normally lack of production assets, knowledge, or
skills (production and entrepreneurship), and thereby they cannot take best advantage
of the loan. Thus in order to boost the macroeconomic effect, improving production
skill and entrepreneur skill is necessary. The average amount of microfinance loan, in
this research, is around 31 million VND which is quite small, thus flexible
determination of the amount of finance may be required in some case. In order to
increase the number of microfinance users, it may be effective to simplify the
procedure and reduce the transaction cost for getting microfinance.
24
Table 8 - Impact of Microfinance (MF) on Output value of selected production Activities
Sugarcane
Other
Livestock Timber Aquaculture Wholesales
Manufacture of Wood
& Furniture
Coefficients from Impact Analysis 1.178 2.203 -1.203 4.999 9.104 6.439
Number of Household using MF in sample 5 200 9 82 4 5
Total Increase in output value in sample 5.9 440.6 -10.8 409.9 36.4 32.2
Estimated Number of HH used MF in rural Vietnama 23,116 924,646 41,609 379,105 18,493 23,116
Total Increase in output value in rural Vietnam 27,230.8 2,036,995.1 -50,055.71 1,895,145.2 168,359.5 148,844.9
Average Amount of MF 47.2 34.9 34.8 35.7 66.3 35.3
Source: Authors' estimation. (Unit in million VND)
a16,384,727 households in rural Vietnam (GSO, 2016) and 3544 in research sample
Table 9 - Macro impact of Microfinance
Output Value
Change in Output
Value Added
Change in VA
No income
linkage
With income
linkage
No income
linkage
With income
linkage
Exogenous sector 180,262,331.9 4,226,519.8 4,226,519.8 175,606,608.7 1,227,576.6 1,227,576.6
Endogenous sector 2,680,853,720.9 4,670,193.7 6,429,386.1 918,635,549.5 1,060,226.2 1,728,121.7
Total 2,861,116,052.8 8,896,713.5 10,655,905.9 1,094,242,158.2 2,287,802.8 2,955,698.3
Source: Authors' estimation. (Unit in million VND)
Table 10 Macro impact of Microfinance by sectors
Output Value
Change in Output
Value Added
Change in VA
No income
linkage
With income
linkage
No income
linkage
With income
linkage
Agriculture, Forestry and Fishery 340,573,962.9 5,514,975.6 5,767,719.3 134,300,317.3 1,750,030.7 1,860,710.7
Mining 123,266,591.3 42,869.3 59,346.3 91,109,499.6 15,534.4 21,862.6
Manufacturing 1,472,059,271.0 2,975,499.1 3,776,322.1 409,489,182.0 322,249.3 472,356.7
Services 925,216,227.6 363,369.5 1,052,518.3 459,343,159.3 199,988.3 600,768.3
Total 2,861,116,052.8 8,896,713.5 10,655,905.9 1,094,242,158.2 2,287,802.8 2,955,698.3
Source: Authors' estimation. (Unit in million VND)
25
5. CONCLUSIONS
This article investigates the impact of microcredit at the disaggregate and
aggregate level. The micro impact of microcredit on households’ income and total output
value from various income sources is conducted using PSM-DID method. The macro
impact is evaluated using input-output analysis.
At micro level, this paper find that microcredit benefit self-employment rather
than activities from agricultural production and common property resources. In particular,
microcredit improves income from self-employment, reduce income from crop and
livestock production while there is no evidence to conclude the impact of microcredit on
income from other activities such as maize or rice production, common property
resources, wage income, etc. In addition, microcredit is found to increase total output
value from self-employment and total earned sources, reduce total output level from
common property resources and there is no significant effect on income from such
sources as wage and agricultural activities.
At macro level, we show output and value added increase 0.37% and 0.27% of
benchmark level respectively due to microfinance. These effect include direct effect and
indirect effect, and show the indirect effect is important as well when we evaluate the
macro effect of microfinance. The microfinance contributes macro economy. In this
study, we focus only on effect of short-run output increase, but microfinance has much
more routes to influence macro economic growth through financial part of the economy.
To evaluate these effects is of interest, and will be done by employing computable
general equilibrium model with real and financial sectors. This is one of our future
studies left for us.
The rural households should be provided with supports so that they can use the
microcredit loan more effectively or improve their income better. Therefore, in addition
to credit, the poor need to be equipped with knowledge and skills in investment in farm/
non-farm activities. Without knowledge and skills, they may not take best advantage of
microcredit or may misuse the loan, and thereby microcredit may result in negligible or
26
even no outcomes. Moreover, rural households, especially the poor, are very risk-averse
and have low resources. Despite their high demand for cash, they do not borrow from any
sources because they may perceive their low ability of repayment. With more knowledge
or skills, they can be more exposed to risks, thereby increasing their demand for more
credit for production or self-employment. Thenceforth, such supplementary supporting
activities as training in entrepreneurial skills or agricultural productions should be
implemented.
REFERENCE
Angelsen, A., & Kaimowitz, D. (1999). Rethinking the causes of deforestation: lessons
from economic models. The world bank research observer, 14(1), 73-98.
Assunçãoa, J., Gandoura, C., Rochaa, R., & Rochab, R. (2013). Does Credit A ect
Deforestation? Evidence from a Rural Credit Policy in the Brazilian
Amazon. Climate Policy Initiative, Rio de Janeiro, Brasil.
Becker, S. O., & Ichino, A. (2002). Estimation of average treatment effects based on
propensity scores. The stata journal, 2(4), 358-377.
Brau, J. C., & Woller, G. M. (2004). Microfinance: A comprehensive review of the
existing literature. The Journal of Entrepreneurial Finance, 9(1), 1.
Buera, F. J., Kaboski, J. P., & Shin, Y. (2012). The macroeconomics of microfinance (No.
w17905). National Bureau of Economic Research (NBER) Working paper #17905,
2012.
Coleman, B. E. (1999). The impact of group lending in Northeast Thailand. Journal of
development economics, 60(1), 105-141.
Coleman, B. E. (2006). Microfinance in Northeast Thailand: Who benefits and how
much?. World development, 34(9), 1612-1638.
Decision No. 306/QD-TTg dated 26 February 2016 of the Prime Minister on adjustment
of loan limits applicable to households conducting production and business
activities in disadvantaged areas.
27
Dehejia, R. H., & Wahba, S. (2002). Propensity score-matching methods for
nonexperimental causal studies. Review of Economics and statistics, 84(1), 151-161.
Demombynes, G., & Vu, L. H. (2015). Demystifying poverty measurement in
Vietnam. World Bank Vietnam Development Economics Discussion Paper, 1.
Dercon, S. (1998). Wealth, risk and activity choice: cattle in Western Tanzania. Journal
of Development Economics, 55(1), 1-42.
Duong, P. B., & Izumida, Y. (2002). Rural development finance in Vietnam: A
microeconometric analysis of household surveys. World Development, 30(2), 319-
335.
Dupas, P., & Robinson, J. (2013). Savings constraints and microenterprise development:
Evidence from a field experiment in Kenya. American Economic Journal: Applied
Economics, 5(1), 163-192.
Fujita, N. (1989). Input-output analysis of agricultural production quotas. The Annals of
Regional Science, 23(1), 41-50.
Godoy, R., O'neill, K., Groff, S., Kostishack, P., Cubas, A., Demmer, J., ... & Martínez,
M. (1997). Household determinants of deforestation by Amerindians in
Honduras. World Development, 25(6), 977-987.
Hulme, D., & Mosley, P. (1996). Finance against poverty (Vol. 2). Psychology Press.
Jodha, N. S. (1992). Common property resources: a missing dimension of development
strategies. World Bank Discussion Paper No. 169, World Bank, Washington, DC.
Karlan, D. S., & Goldberg, N. (2007). Impact evaluation for microfinance: Review of
methodological issues. World Bank, Poverty Reduction and Economic Management,
Thematic Group on Poverty Analysis, Monitoring and Impact Evaluation.
Khandker, S. R. (1998). Fighting poverty with microcredit: experience in Bangladesh.
Published for the World Bank. Oxford University Press, pp. xii + 228 pp.
Khandker, S. R., & Faruqee, R. R. (2003). The impact of farm credit in
Pakistan. Agricultural Economics, 28(3), 197-213.
Khandker, S. R. (2010). Handboook on Impact Evaluation – Quantitative Method and
Practice. The World Bank, Development Economics.
28
Khandker, S. R., & Koolwal, G. B. (2015). How has microcredit supported agriculture?
Evidence using panel data from Bangladesh. Agricultural Economics.
Khoi, P. D., Gan, C., Nartea, G. V., & Cohen, D. A. (2013). Formal and informal rural
credit in the Mekong River Delta of Vietnam: Interaction and accessibility. Journal
of Asian Economics, 26, 1-13.
Lensink, R. and Pham, T. T. T. (2012). The impact of microcredit on self-employment
profits in Vietnam. Economics of Transition, 20, 73–111. Doi: 10.1111/j.1468-
0351.2011.00427.x
Lester, R. A. (1946). Shortcomings of Marginal Analysis for the Wage-Employment
Problems. American Economic Review, 36, 63-82.
Li, X., Gan, C. & Hu, B. (2011). The welfare impact of microcredit on rural households
in China, The Journal of Socio-Economics, 40(4), 404-411, DOI:
10.1016/j.socec.2011.04.012.
Mahjabeen, R. (2008). Microfinancing in Bangladesh: Impact on households,
consumption and welfare. Journal of Policy modeling, 30(6), 1083-1092.
Maksudova, N. (2010). Macroeconomics of Microfinance: How Do the Channels Work?.
Working paper #423, Center for Economic Research and Graduate Education,
Charles University, 2010.
Morduch, J., & Haley, B. (2002). Analysis of the effects of microfinance on poverty
reduction. New York: NYU Wagner Working Paper, 1014.
Nguyen, V. C. (2008). Is a governmental micro-credit program for the poor really pro-
poor? Evidence from Vietnam. The Developing Economies, XLVI(2), 151-187.
Phan Dinh Khoi, Gan, C., Nartea, G. V. & Cohen, D. A. (2014). The impact of
microcredit on rural households in the Mekong River Delta of Vietnam. Journal of
the Asia Pacific Economy, 19(4), 558-578, DOI: 10.1080/13547860.2014.920591
Raihan, S., Osmani, S. R., & Khalily, M. B. (2017). The macro impact of microfinance in
Bangladesh: A CGE analysis. Economic Modelling, 62, 1-15.
29
Reis, N. & Mollinga, P. P (2012). Water Supply or ‘Beautiful Latrines’? Microcredit for
Rural Water Supply and Sanitation in the Mekong Delta, Vietnam. ASEAS –
Austrian Journal of South-East Asian Studies, 5(1), 10-29.
Rosenbaum, P. R. & Rubin, D. B. (1983). The Central Role of the Propensity Score in
Observational Studies for Causal Effects. Biometrika, 70, 41-55.
Takahashi, K., Higashikata, T., & Tsukada, K. (2010). THE SHORT‐TERM POVERTY
IMPACT OF SMALL‐SCALE, COLLATERAL‐FREE MICROCREDIT IN
INDONESIA: A MATCHING ESTIMATOR APPROACH. The Developing
Economies, 48(1), 128-155.
Vilhelm, V., Spicka, J., & Valder, A. (2015). Public Support of Agricultural Risk
Management-Situation and Prospects. AGRIS on-line Papers in Economics and
Informatics, 7(2), 93.
Villa, J.M., 2016. diff: Simplifying the estimation of difference-in-differences treatment
effects. Stata Journal, 16, 52-71.
Yunus, M. (2003). Banker to the poor: The story of the Grameen Bank. Aurum Press
Limited.
